Targeting tumor-associated macrophages using mRNA lipid nanoparticles for cytotoxic T lymphocyte–mediated cancer immunotherapy

Abstract

The immunosuppressive tumor microenvironment (TME) remains one of the main obstacles that limit responsiveness to immunotherapy. Recently, lipid nanoparticles carrying messenger RNA (mRNA) have emerged as a promising strategy to modulate the immunosuppressive TME, with the ultimate goal of sustaining anticancer immunity of cytotoxic T lymphocytes (CTLs). However, there are challenges with mRNA-based cytokine/chemokine therapies such as the lack of specific targeting, low therapeutic efficacy, and elevated toxicity, raising concerns about their clinical translation. Here, we have developed an antibody-coated lipid nanoparticle (Ab-LNP) delivery system that targets the protein triggering receptor expressed on myeloid cells 2 (TREM2) expressed by tumor-associated macrophages (TAMs). We demonstrate that codelivery of a Toll-like receptor agonist and CXCL9-encoding mRNA encapsulated in our Ab-LNP successfully ameliorates immunosuppression and improves tumor infiltration and activity of CTLs. Further combination with immune checkpoint inhibitors against PD-L1 and CTLA-4 promoted a CTL-favoring immunological environment and durable memory immunity. Our Ab-LNPs targeting TAMs highlight the potential of reprogramming immunosuppressive TME to enhance CTL activity for improved response to cancer immunotherapy.

INTRODUCTION

Cancer immunotherapy has revolutionized cancer treatment and paved the way for the development of new translatable therapies (1). Among them, therapeutic modalities that modulate the antitumor response of cytotoxic T lymphocytes (CTLs; or CD8⁺ T cells) have been identified as one of the most effective strategies (2–4), for example, immune checkpoint blockade (ICB) (5, 6) and chimeric antigen receptor (CAR) T cell therapy (7, 8). Despite extensive efforts, clinical outcomes across different cancers, particularly solid tumors, remain limited mainly due to the immunological heterogeneity of the tumor microenvironment (TME) (9). In particular, patient responses to ICB are highly dependent on the immunotype of the tumors (10), which limits clinical benefit to many patients with solid cancer (11). Furthermore, the immunosuppressive TME can induce T cell exhaustion and impede tumor penetration of CAR T cells, thereby compromising their therapeutic efficacy in treating solid tumors (12). Therefore, poor immune infiltration and an immunosuppressive TME remain major barriers to effective patient responses to cancer immunotherapy.

To overcome these barriers for CTL-mediated immunotherapy, targeting suppressive immune cells and promoting CTL-favorable immune signatures in the TME have become a promising strategy in developing new cancer immunotherapy (13, 14). As one of the most abundant cell types in tumors, tumor-associated macrophages (TAMs) orchestrate an immunosuppressive TME by modulating cytokine and chemokine secretion, expressing inhibitory immune checkpoints, and suppressing antigen-presenting cells (APCs) and T cell activities, which often correlates with a poor prognosis in many patients with cancer (15). Therefore, targeting TAMs has garnered considerable research interest, with a focus on TAM depletion using antibodies and/or small-molecule inhibitors, inhibiting TAMs recruitment (e.g., blocking CCL2/CCR2 signaling pathway), and reprogramming TAMs to an immune-stimulatory phenotype and activating their phagocytic capacity against tumor cells (16–18). Given that TAMs naturally produce a broad range of chemokines, their role within the tumor chemokine network has been recently reported (19, 20). This opens a promising yet unexplored avenue: genetically modifying TAMs to enhance chemokine secretion and thereby stimulate immune cells such as CTLs in TME for cancer immunotherapy.

The specific delivery of payloads to TAMs using nanoparticles (NPs) has been demonstrated in the past (21). Lipid NPs (LNPs) represent one of the most promising delivery vehicles for therapeutic mRNAs, as demonstrated by the efficacy and safety of COVID-19 mRNA-LNP vaccines (22). Various aptamer-, peptide-, and antibody-modified LNPs have been developed for improved targeting (23–25). However, targeting immune cells in TMEs remains challenging due to different factors, including the heterogeneity of solid tumors, competitive uptake by nontarget cells, and physical barriers affecting tissue penetration.

In this study, we developed a previously unreported antibody-coated LNP (Ab-LNP) delivery system coloaded with the Toll-like receptor 7/8 (TLR7/8) agonist resiquimod and mRNA encoding the chemokine C-X-C motif ligand 9 (CXCL9) to inhibit TAM immunosuppression and enhance tumor infiltration of CTLs for cancer immunotherapy (Fig. 1). By coating LNPs with an antibody against triggering receptor expressed on myeloid cells 2 (TREM2) expressed on the surface of TAMs (26), we showed improved targeting of Ab-LNP to TAMs as compared to 4T1 tumor cells in vitro and in vivo. Furthermore, codelivery of resiquimod and CXCL9 mRNA by Ab-LNPs simultaneously reprogrammed immunosuppressive TAMs and promoted the secretion of CXCL9 as a chemoattractant for CTLs. Further in vivo studies of the Ab-LNPs validated their ability to inhibit the immunosuppressive phenotypes of TAMs and the expression of CXCL9 in the TME, stimulating infiltration and activity of CTLs. Moreover, the results revealed that the Ab-LNPs could augment the immunomodulating effect of immune checkpoint inhibitors and create a CTL-favorable immunological environment. Overall, this work developed a paradigm of an Ab-LNP system that targets and engineers immunosuppressive TAMs to sustain CTLs in solid tumor for improving responsiveness to cancer immunotherapy.

Fig. 1. Schematic illustration of TAM-targeted resiquimod mRNA-CXCL9 LNPs for cancer immunotherapy.

(1) TREM2 Ab-LNPs target surface markers of TAMs to improve uptake efficacy and restrict their interaction with tumor cells in the TME. (2) Intracellular delivery of TLR7/8 agonists binds endosomal TLR receptors and reprograms immunosuppressive TAMs to proinflammatory macrophages. The codelivery LNP system achieved effective production of CXCL9 chemokine. (3) Simultaneous reduction of immunosuppressive TAMs and production of the chemokine CXCL9 creates a favorable immunological environment for CTLs and attracts their trafficking to tumors via the CXCL9-CCR3 axis. (4) Combination of the LNPs and immune checkpoint inhibitors improves the presence of immunostimulatory T helper (T_H) cells and CTLs and promotes central memory immunity of CTLs in the tumor-draining lymph node (TdLN) simultaneously. Created in BioRender. Zhang, Y. (2026), https://BioRender.com/8qowqry.

RESULTS

Identification of chemokine CXCL9 as a regulator of CTLs in breast cancer

Chemokines in the TME not only influence tumor cell proliferation but also regulate immune cell migration and activity, thereby contributing to shaping the immunological phenotype of solid tumors. To identify the key chemokines responsible for intratumor infiltration of CTLs (i.e., CD8⁺ T cells), we first investigated the correlation between mRNA levels of chemokines and CD8A gene expression in patients with breast cancer using the LinkedOmics platform (fig. S1A) (27). Among all types of CXCL chemokines that were positively correlated with CD8A level, CXCL9 exhibits a substantially high correlation coefficient of 0.8191 (fig. S1B), suggesting that it may play a role in facilitating CTL infiltration in human breast cancer. In agreement with this, a Kaplan-Meier survival curve taken from The Cancer Genome Atlas Program (TCGA) database shows a significantly longer survival for patients with breast cancer with high CXCL9 expression compared with those with low expression (fig. S1C).

To understand the regulatory functions of CXCL9 in breast cancer and determine whether they are relevant to immunity, we used a gene set enrichment analysis tool to unravel 10 biological pathways with the highest and lowest normalized enrichment scores. Gene set enrichment analysis reveals a significant positive association between elevated CXCL9 expression and enhanced lymphocyte immune responses and adaptive immune functions in breast cancer (P < 0.0001, fig. S1D). In addition, we examined the expression of 68 cancer immune signatures in relation to CXCL9 expression and found that CXCL9 expression correlated with stimulatory genes associated with CD8⁺ T activity, such as granzyme A and interferon-γ (IFN-γ) (fig. S1, E and F). Therefore, we hypothesize that restoring CXCL9 expression in breast tumors could enhance CD8⁺ T infiltration and activity, thereby improving anticancer immune responses.

Design and characterization of Ab-LNP delivery systems

mRNA LNPs were synthesized by mixing an mRNA aqueous solution with lipids in ethanol as previously described (28). Ab-LNPs were prepared by a two-step process (Fig. 2A). Maleimide-functionalized resiquimod mRNA (RmRNA) LNPs were prepared by mixing mRNA and resiquimod in an aqueous solution and a lipid mixture in ethanol (29), followed by coating with TREM2 monoclonal antibodies using thiol-maleimide bioconjugation to form Ab-LNPs (T-RmRNA LNPs) (30). In addition, we benchmarked our T-RmRNA LNPs against LNPs targeting the mannose receptor of TAMs (M-RmRNA LNPs) by replacing maleimide-PEG with mannose-PEG (31). We summarized abbreviations for all the formulations of LNPs used in this study (table S1).

Fig. 2. Optimized coating of TREM2 antibody on LNPs facilitates mRNA delivery to immunosuppressive macrophages with improved specificity in vitro.

(A) Design of mRNA and resiquimod codelivered LNPs, TREM2 Ab-LNPs, and benchmarking mannose-modified LNPs. Lipids in ethanol are mixed with mRNA and resiquimod in 50 mM citrate buffer (pH 3.0), followed by dialysis in phosphate-buffered saline (PBS) for 24 hours. Processed TREM2 antibodies were conjugated to maleimide groups on the surface of LNPs. Created in BioRender. Zhang, Y. (2026), https://BioRender.com/30d5utt. (B) TREM2 antibody–coating efficiencies with different molar ratios of input antibody and maleimide on LNPs. (C) Hydrodynamic size, polydispersity, and (D) zeta potentials of the LNPs. (E) Transmission electron microscopy images of LNPs and Ab-LNPs with optimized coating efficiency [scale bars, 200 nm and (inset) 25 nm]. (F) TREM2 and MRC1 expression in mIL-4–treated (immunosuppressive), LPS + mIFN-γ–treated (proinflammatory) BMDMs and 4T1 tumor cells. (G) DiD-labeled LNPs uptake by immunosuppressive BMDMs after incubation for 1, 4, and 24 hours (*P < 0.05). (H) Green fluorescence protein (GFP) signal of immunosuppressive BMDMs treated with the mRNA LNPs for 24 hours (n = 3 per group; *P < 0.05 and **P < 0.01). (I) GFP signal of 4T1 tumor cells treated with the mRNA LNPs for 24 hours (n = 3 per group; **P < 0.01 and ****P < 0.0001). Cells were treated with mRNA LNPs at an mRNA concentration of 0.5 μg/ml.

First, we optimized the loading of mRNA and resiquimod in the RmRNA LNPs (fig. S2), as well as TREM2 antibody (αTREM2) coating. At the molar ratio of 1:1 (antibody to maleimide), the highest antibody-coating efficiency of 91.6% was achieved (Fig. 2B). As characterized by dynamic light scattering (DLS), hydrodynamic sizes and polydispersity of RmRNA LNPs, M-RmRNA LNPs, and T-RmRNA LNPs (antibody-to-maleimide molar ratio of 1:1) ranged from 80 to 100 nm and 0.096 to 0.11 (Fig. 2C), respectively. When increasing antibody coating to 2:1 and 4:1, both hydrodynamic size and polydispersity of T-RmRNA LNPs significantly increased. There was no obvious change in zeta potential for T-RmRNA LNPs as compared to RmRNA LNPs and M-RmRNA LNPs had (Fig. 2D). The size and morphology of RmRNA LNPs and T-RmRNA LNPs were also characterized by transmission electron microscopy (Fig. 2E and fig. S3), showing consistent results as DLS.

In vitro uptake specificity of Ab-LNPs by immunosuppressive macrophages

Next, we examined the in vitro uptake of the LNPs using immunosuppressive macrophages and 4T1 tumor cells to assess their targeting capabilities. To generate proinflammatory and immunosuppressive macrophages, bone marrow–derived macrophages (BMDMs) from healthy BALB/c mice were polarized using mouse interleukin-4 (mIL-4) or lipopolysaccharide (LPS) and mouse IFN-γ (mIFN-γ), respectively, as previously reported (32). Successful polarization of BMDMs to these two phenotypes was confirmed using CD86 and CD206 and by quantifying mRNA levels of immunosuppressive markers: arginase 1 (ARG1) and mannose receptor C-type 1 (MRC1), and proinflammatory markers: nitric oxide synthase 2 (NOS2) and interleukin-12B (IL-12B) (fig. S4). Further analysis showed that TREM2 and MRC1 were highly expressed by immunosuppressive macrophages compared with proinflammatory macrophages and 4T1 tumor cells (Fig. 2F), indicating that targeting these two surface markers is a promising strategy to achieve TAM-specific delivery.

To assess cell targeting via antibody binding and to exclude any cellular signaling effects from resiquimod, we used LNPs encapsulating only mRNA-encoding enhanced green fluorescence protein (mRNA-eGFP) as a reporter in cell assays. We first labeled the LNPs with the fluorescent dye 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindodicarbocyanine, 4-chlorobenzenesulfonate (DiD) as previously described (33) followed by incubation with mIL-4–treated BMDMs for 1, 4, and 24 hours, and the cellular uptake of LNPs was evaluated by fluorescence microscopy (Fig. 2G and fig. S5). Quantification of cellular DiD signal normalized to the cell number showed that mannose- and TREM2-coating strategies significantly improved LNP uptake at 4 and 24 hours, respectively. Furthermore, mRNA delivery efficacy by the LNPs was characterized by incubating mRNA-eGFP LNPs (mRNA LNPs), isotype control Ab-LNPs (iso-mRNA LNPs), TREM2 Ab-LNPs (T-mRNA LNPs), and mannose-modified LNPs (M-mRNA LNPs) with mIL-4–treated BMDMs and 4T1 tumor cells, followed by flow cytometry analysis of their green fluorescence protein (GFP) signal. While the M-mRNA LNPs showed the highest GFP signal for both mIL-4–treated BMDMs and 4T1 tumor cells, T-mRNA LNPs exhibited increased uptake in mIL-4–treated BMDMs and a decrease in 4T1 tumor cells (Fig. 2, H and I). Therefore, our TREM2 antibody–coating strategy demonstrated effective improvement of in vitro mRNA delivery to mIL-4–treated BMDMs and reduced mRNA delivery to 4T1 tumor cells as compared to nonmodified LNPs.

In vivo targeting of Ab-LNPs to TAMs

The in vivo biodistribution of mRNA LNPs, T-mRNA LNPs, and M-mRNA LNPs was assessed in the orthotopic 4T1 breast cancer model. Firefly luciferase–encoding mRNA (mRNA-FLuc) and DiD dye were coencapsulated in LNPs to evaluate mRNA delivery and track biodistribution of LNPs, respectively. A single dose of mRNA LNPs, T-mRNA LNPs, or M-mRNA LNPs containing 9 μg of mRNA-FLuc, along with phosphate-buffered saline (PBS) as a control, was intravenously injected into 4T1 tumor-bearing mice. After 12 hours, various tissues, including the tumor, were collected, followed by quantification of the luminescence (Fig. 3, A and B) and fluorescence signal (Fig. 3, C and D). This analysis showed that mouse liver, spleen, and lung were the main organs with high-level accumulation of LNPs, which is in line with other studies (34, 35). Moreover, T-mRNA LNPs and M-mRNA LNPs showed a trend toward enhanced tumor localization for both mRNA and LNPs as compared to the control mRNA LNPs without any targeting moiety. This result indicated that TREM2 antibody and mannose modification did not have much impact on tumor and organ distribution of LNPs.

Fig. 3. TREM2 Ab-LNPs enable specific uptake by TAMs in 4T1 solid tumors.

(A) Bioluminescence images of 4T1 tumor-bearing mice, excised tumors and organs at 12 hours after injection of PBS, DiD-labeled mRNA-FLuc LNPs, TREM2 antibody–coated DiD-labeled mRNA-FLuc LNPs (T-mRNA LNPs), and mannose DiD-labeled mRNA-FLuc LNPs (M-mRNA LNPs; n = 3 per group). Each mouse received PBS or mRNA-FLuc LNP treatment via intravenous injection with a single dose of mRNA-FLuc at 0.45 mg/kg mouse body weight. Biodistribution was assessed at 12 hours postinjection. (B) Quantitative luminescence of tumors and organs to indicate mRNA-FLuc delivery by the LNPs (n = 3 per group; *P < 0.05). Signal detected from the PBS group was regarded as background luminescence. (C) Fluorescence images of the mice, excised tumors, and organs. Ex vivo fluorescence imaging was performed at 620 nm (excitation) and 670 nm (emission). (D) Quantitative fluorescence of DiD in tumors and organs to indicate distribution of the LNPs (n = 3 per group; *P < 0.05). Signal detected from PBS group was regarded as background fluorescence. DiD fluorescence intensity of (E) TAMs, (F) 4T1 tumor cells, and (G) their ratio was characterized by flow cytometry (n = 3 per group; **P < 0.01, ***P < 0.001, and ****P < 0.0001). (H) Quantitative analysis of colocalization of DiD-labeled LNPs and F4/80⁺ macrophages in mouse tumors characterized by immunofluorescence imaging (n = 3 per group). (I) Representative immunofluorescence images of the tumor tissues. Blue, nuclei; green, F4/80⁺ macrophages; red, DiD-labeled LNPs. Scale bars, 50 μm.

To gain a better understanding of the cellular uptake of the LNPs in tumor cells and TAMs, we processed the 4T1 tumors and assessed DiD fluorescence intensity in tumor cells and TAMs by flow cytometry. T-mRNA LNPs showed a significantly enhanced uptake by macrophages compared with mRNA LNPs and M-mRNA LNPs (Fig. 3E), but with reduced tumor uptake compared with M-mRNA LNP (Fig. 3F). Notably, T-mRNA LNPs showed a significantly higher uptake ratio of macrophages to tumor cells (ratio of 5.11 ± 0.47) as compared to mRNA LNPs (ratio of 2.77 ± 0.1) and M-mRNA LNPs (ratio of 2.91 ± 0.23), revealing a significant TAM-specific uptake in vivo (Fig. 3G).

We also processed the tumor tissues for immunofluorescence imaging to characterize the colocalization of DiD⁺ cells with LNP uptake and F4/80⁺ macrophages per mouse tumor (Fig. 3H). Quantification of the percentage tumor area with dual DiD and F4/80⁺ staining showed enhanced uptake of T-mRNA LNPs by macrophages (Fig. 3I). Consistent with our in vitro results, T-mRNA LNPs achieved effective in vivo accumulation in TAMs, outperforming the previously reported strategy targeting the mannose receptor on TAMs.

Codelivery of resiquimod improves mRNA translation efficacy in macrophages

Next, we tested the effect of LNPs with different loading amounts of mRNA-encoding CXCL9 (mRNA-CXCL9) and resiquimod on immunosuppressive BMDMs. We determined the optimal combination of 2.53 to 2.95 wt % mRNA-CXCL9 and 0.11 to 0.54 wt % resiquimod for chemokine secretion and reprogramming macrophages, respectively (fig. S6). LNP with 2.95 ± 0.09 wt % of mRNA-CXCL9 and 0.19 ± 0.06 wt % of resiquimod showed the best performance in the production of CXCL9, down-regulation of ARG1, and NOS2 up-regulation. Therefore, we used this ratio to prepare resiquimod and mRNA-coloaded LNPs (RmRNA LNPs) for further in vitro and in vivo studies.

To understand the impact of resiquimod on intracellular mRNA translation, we assessed the cellular uptake and mRNA delivery of mRNA LNPs and RmRNA LNPs and separated delivery of mRNA LNPs and resiquimod in immunosuppressive BMDMs using LNPs coloaded with DiD and mRNA-eGFP. mIL-4–pretreated BMDMs were incubated with PBS, mRNA LNPs, RmRNA LNPs, or a mixture of mRNA LNPs and resiquimod (mRNA LNPs + R) for 4 and 24 hours, followed by staining with the lysosome marker LysoTracker Red and 4′,6-diamidino-2-phenylindole (DAPI) and analyzed by confocal microscopy to compare lysosomal localization of LNPs, LNP uptake, and mRNA translation under different conditions (Fig. 4, A and B). At 4 hours (Fig. 4A), there was no difference in colocalization of DiD and lysosomes (Fig. 4Ca). Both the RmRNA LNP and mRNA LNP + R groups showed a significant increase in DiD signal (Fig. 4Cb), indicating enhanced cellular uptake of LNPs, while only mRNA LNP + R showed a significant increase in GFP intensity (Fig. 4Cc). At 24 hours (Fig. 4B), the colocalization area of DiD and lysosomes in RmRNA LNPs (0.64%) exhibited a substantially lower percentage than mRNA LNPs (1.18%) and mRNA LNPs + R (1.19%, Fig. 4Da). Meanwhile, it must be noted that the DiD signal of RmRNA LNPs was comparable to mRNA LNPs (Fig. 4Db), but its GFP signal was significantly higher than mRNA LNPs (Fig. 4Dc). These results showed that treatment with free resiquimod molecules can improve LNP uptake and facilitate fast intracellular mRNA delivery. This enhancement of LNP uptake is hypothetically ascribed to activation of phagocytic activity after free resiquimod molecules bind to TLR7 on the cell surface of macrophages (36). In comparison, coencapsulation and delivery via RmRNA LNPs did not increase total LNP uptake by activating TLRs on the cell surface. The LNPs deliver resiquimod to endosomal TLR7 and TLR8 and potentially improved mRNA translation efficacy through decreasing further entrapment of LNPs in lysosomes (37).

Fig. 4. Codelivery of resiquimod and mRNA-CXCL9 reprogram macrophages and stimulate their CXCL9 secretion efficiently in vitro.

(A and B) Evaluation of cell uptake, lysosomal entrapment, and mRNA translation of mRNA-eGFP (mRNA) LNPs, resiquimod mRNA-eGFP–coloaded (RmRNA) LNPs, and separate treatment of mRNA LNPs and resiquimod at incubation times of 4 and 24 hours using confocal laser scanning microscopy. LNPs were labeled with DiD and incubated with immunosuppressive BMDMs at an mRNA concentration of 0.5 μg/ml. PBS treatment was used to detect background signal. Cells were stained with 4′,6-diamidino-2-phenylindole (DAPI) and LysoTracker Red (Lyso Red), and, then, images under DAPI, APC, and FITC channels were obtained and merged. Scale bars, 10 μm. (C) (a) Colocalization analysis, (b) DiD intensity, and (c) GFP intensity of cells after 4 hours of the LNPs; (D) (a) Colocalization analysis, (b) DiD intensity, and (c) GFP intensity of cells after 24 hours of incubation with the LNPs (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). (E) Schematic diagram of in vitro chemokine secretion and macrophage polarization experiment. Created in BioRender. Zhang, Y. (2026), https://BioRender.com/7wkdh78. M-CSF, macrophage colony-stimulating factor. (F) CXCL9 secretion in cell medium after incubation with RmRNA LNPs, TREM2 antibody–coated RmRNA LNPs (T-RmRNA LNPs), and mannose RmRNA LNPs (M-RmRNA LNPs) for 24 hours. (G) ARG1 and NOS2 gene expression of the cells after incubation with the LNPs. Gene expression was normalized by the expression in the BMDMs + mIL-4 group (**P < 0.01 and ****P < 0.0001).

We next assessed chemokine secretion and macrophage-reprogramming in vitro using resiquimod and mRNA-CXCL9-coloaded LNPs (RmRNA LNPs), TREM2 antibody–coated RmRNA LNPs (T-RmRNA LNPs), and mannose-modified LNPs (M-RmRNA LNPs). Immunosuppressive BMDMs were incubated with these three types of LNPs for 24 hours. After replacing with fresh cell culture medium and incubating for another 24 hours, CXCL9 production was quantified by enzyme-linked immunosorbent assay (ELISA), and the cells were harvested for assessment of ARG1 and NOS2 gene expression (Fig. 4E). Treatment with T-RmRNA LNPs produced the highest concentration of secreted CXCL9 (Fig. 4F), and the cells were successfully transformed to their proinflammatory subsets, as evidenced by significant 2.5-fold down-regulation of ARG1 and 89.5-fold up-regulation of NOS2 (Fig. 4G). By direct comparison with M-RmRNA LNPs, T-RmRNA LNPs represent a more effective strategy to simultaneously engineer chemokine production and regulate immunosuppression of macrophages.

In vivo anticancer therapeutic and immunological effects of the LNPs

Next, we investigated the therapeutic efficacy and immunological effects of T-RmRNA LNPs in the orthotopic 4T1 tumor model. We focused on primary 4T1 solid tumor progression at early-to-middle stage to avoid the impact of tumor ulceration on animal well-being and immune profiling of the tumor. First, we compared mRNA LNPs, RmRNA LNPs, and T-RmRNA LNPs to distinguish the different effects of mRNA-CXCL9, resiquimod, and TREM2 antibody coating of LNPs. After tumor establishment, three doses of PBS or the LNPs were injected intravenously on days 7, 11, and 14 (12 μg of mRNA per dose; Fig. 5A). All LNP treatment groups showed a decrease in tumor volume compared with the PBS control group and were well tolerated with minimal impact on mice body weights (Fig. 5B). Correspondingly, as compared to the PBS group, there was a significant but moderate decrease in tumor weight for mice treated with T-RmRNA LNPs, whereas mRNA LNPs and RmRNA LNPs only showed a trend toward tumor growth inhibition (Fig. 5C). In addition, CXCL9 levels were elevated in tumor tissues following treatment with LNPs, particularly in the T-RmRNA LNP group that showed a significant fourfold increase in CXCL9 compared with the control (Fig. 5D). Consistent with the in vitro result, mRNA delivery by T-RmRNA LNPs successfully modulated chemokine levels within the TME.

Fig. 5. TREM2 antibody–coated, resiquimod, and mRNA-CXCL9–coloaded LNPs suppress 4T1 tumor growth in vivo by reducing immunosuppressive TAMs, boosting tumor infiltration and activity of CTLs.

(A) Treatment schedule of mRNA-CXCL9 LNPs, RmRNA LNPs, and T-RmRNA LNPs. One million 4T1 tumor cells were inoculated subcutaneously into the fourth mammary fat pad of a BALB/c mouse on day 0. The mRNA LNPs were injected intravenously into the mouse tail vein at an equivalent amount of 12 μg of mRNA-CXCL9 on days 7, 11, and 14. PBS treatment was set as the control group for comparison. Created in BioRender. Zhang, Y. (2026), https://BioRender.com/44td5ty. iv, intravenous. (B) Tumor volume and body weight of the mice receiving LNP treatment. (C) Tumor weight at the endpoint of the study (n = 5 per group; *P < 0.05). (D) CXCL9 level in 4T1 tumor with treatment of PBS, mRNA LNPs, RmRNA LNPs, and T-RmRNA LNPs (n = 5 per group; *P < 0.05). (E) ARG1 and NOS2 expression of tumors with treatment of PBS, mRNA LNPs, RmRNA LNPs, and T-RmRNA LNPs (n = 5 per group; *P < 0.05). (F) Percentages and (G) concatenated dot plots of immunosuppressive and proinflammatory TAMs in the tumors after LNP treatment (n = 5 per group; *P < 0.05). (H) Number of CTLs, i.e., CD8⁺ T cells infiltrated per gram of tumors (n = 5 per group; *P < 0.05 and **P < 0.01). (I) CD44 expression of tumor-infiltrated CD4⁺ T cells and CD8⁺ T cells (n = 5 per group; *P < 0.05). (J) Number of IFN-γ⁺ spots and their representative images from immune cells in TdLNs (n = 5 per group; *P < 0.05).

To understand the immunological impacts of the LNP treatment, particularly on dendritic cells (DCs), TAMs, and CD8⁺ T cells, we assessed immunotypes of the 4T1 tumor tissues after treatment. As it has been reported that TLR7/8 agonists such as resiquimod are capable of activating APCs such as DCs (38), we evaluated whether resiquimod-loaded LNPs could influence DC activity in 4T1 tumors by evaluating major histocompatibility complex class II (MHC-II) and CD86 expression of DCs, corresponding to their antigen-presenting activation and maturation (39). Profiling immune cells in the solid tumor indicated that mRNA LNPs, RmRNA LNPs, and T-RmRNA LNPs can improve immune cell infiltration in 4T1 tumors, including CD4⁺ T cells, DCs, and macrophages. It is noted that a significant decrease of other CD11b⁺ cells (i.e., CD11b⁺ cells excluding macrophages) was observed for the T-RmRNA LNP group (fig. S7A). No significant changes in MHC-II and CD86 levels in DCs were observed between the T-RmRNA LNP treatment group and the PBS control, which may be attributed to the low loading content of resiquimod (fig. S7B). Next, we examined ARG1 and NOS2 levels in the tumors to evaluate functional markers of TAMs in TME (Fig. 5E). As compared to the PBS group, T-RmRNA LNPs can significantly decrease ARG1 and increase NOS2 in the tumor, which demonstrated successful regulation of immunosuppressive functions and stimulation of inflammatory effects. To further assess immunological phenotypes of TAMs, we gated immunosuppressive and proinflammatory subsets based on CD86 and CD206 expression (Fig. 5, F and G). Compared with PBS control, T-RmRNA LNP treatment resulted in a significant 63.2% reduction in CD206⁺CD86⁻ immunosuppressive TAMs but not affected CD206⁻CD86⁺ proinflammatory TAMs, indicating successful regulation of immunosuppression in TAMs.

Analysis of CTLs showed an increase in the number of CD8⁺ T cells infiltrating into 4T1 tumors following all LNP treatments (Fig. 5H), with notable increases of 68.5 and 26.2% observed for the mRNA LNPs and T-RmRNA LNPs, respectively, compared with the PBS control. This result supported our hypothesis that reexpressing CXCL9 via mRNA LNPs could promote CD8⁺ T cells trafficking to tumors. In addition, we examined CD44 expression as an activation marker of T lymphocytes localized at tumor tissues. T-RmRNA LNP treatment resulted in elevation of CD44 expression in both CD8⁺ T cells (27.6%) and CD4⁺ T cells (19.9%), demonstrating their capability of activating T cells in tumors (Fig. 5I).

To further analyze the effect of these treatments on T cell activity and antitumor response, we used an enzyme-linked immunosorbent spot (ELISpot) assay to evaluate IFN-γ release under stimulation with tumor lysates and resiquimod. We isolated immune cells from tumor-draining lymph node (TdLN; left inguinal lymph node) and mesenteric lymph node (mLN) from tumor-bearing mice and then stimulated the cells with 4T1 tumor lysate and resiquimod to titrate the concentrations of stimulants. Cells treated with anti-CD3ε were used as the positive control. As evidenced by IFN-γ⁺ spot numbers, immune cells in TdLN responded well to the stimulants, whereas those in mLN showed neglectable response (fig. S8). Using an IFN-γ⁺ ELISpot assay, we found that T-RmRNA LNP treatment resulted in a significantly higher number of IFN-γ⁺ spots compared with the PBS control, indicating excellent antitumor activity of immune cells, mainly cytotoxic CD8⁺ T cells (Fig. 5J). Together, our results demonstrated that T-RmRNA LNPs enhanced the tumor infiltration and activity of CD8⁺ T cells, subsequently contributing to a reduction in 4T1 tumor growth via CTL-mediated immunotherapy.

Combination of T-RmRNA LNPs and immune checkpoint inhibitors

Given that our T-RmRNA LNPs increased CXCL9 tumor expression, reduced TAM-mediated immunosuppression, and promoted CD8⁺ T tumor infiltration and activity, we next explored combining T-RmRNA LNPs with immune checkpoint inhibitors with the hypothesis that T-RmRNA LNPs could alter immunotypes of 4T1 solid tumors to improve the efficacy of ICB therapy. Based on previous findings, 4T1 tumors are known for a TME dominated by suppressive myeloid cells, leading to a limited response to the single modality of ICB (40). In addition, our immune signature analysis showed that CXCL9 expression in patients with breast cancer highly correlated to expression of two inhibitory immune checkpoints: PDCD1 (PD-1) and CTLA-4 (fig. S1F). Therefore, we chose anti–PD-L1 and anti–CTLA-4 antibody treatment to combine with T-RmRNA LNPs, aiming to regulate PD-1/PD-L1 through blocking PD-L1 that is abundantly expressed by myeloid cells and to suppress CXCL9-associated inhibitory checkpoints, thereby improving the outcomes of CTL immunotherapy in TNBC.

Combination immunotherapy consisting of LNPs and immune checkpoint inhibitors was performed in the orthotopic 4T1 mouse tumor model. Tumor-bearing mice received three doses of intravenous injection of vehicle control LNPs (i.e., LNPs without resiquimod and mRNA-CXCL9) or T-RmRNA LNPs in combination with isotype control antibodies (isotype monoclonal antibody) or PD-L1 and CTLA-4 antibodies (αPD-L1 αCTLA-4, Fig. 6A). Control mice received PBS. Tumor volume and mice body weight were monitored during treatment. Unexpectedly, treatment with vehicle LNPs + αPD-L1 αCTLA-4 did not inhibit 4T1 tumor growth but showed a trend of increase in tumor volume at day 9. We noted that both the T-RmRNA LNPs + isotype antibody and the T-RmRNA LNPs + αPD-L1 αCTLA-4 treatment groups showed significant decreases in tumor volume as compared to the vehicle LNPs group, but there is no significant difference between those two groups (Fig. 6B). This was also reflected in the tumor weights taken at the end of the experiment (Fig. 6C). Therefore, primary 4T1 tumor did not show therapeutic respond to αPD-L1 αCTLA-4 treatment alone in our study, whereas combination of T-RmRNA LNPs and αPD-L1 αCTLA-4 treatment could maintain a significant therapeutic efficacy as well as the T-mRNA LNPs but did not further inhibit tumor growth.

Fig. 6. Combination with immune checkpoint inhibitors anti–PD-L1 and anti–CTLA-4 facilitates a favorable immune environment for CTLs in tumors.

(A) Treatment schedule of vehicle LNPs and T-RmRNA LNPs in combination with anti–PD-L1 and anti–CTLA-4 or their isotype control antibodies. The LNPs carrying 4 μg of mRNA were injected intravenously to mouse tail vein on days 4, 6, and 8 after 4T1 tumor inoculation. Immune checkpoint inhibitors anti–PD-L1 (100 μg per dose) and anti–CTLA-4 (100 μg per dose) or their isotype control antibodies were injected intraperitoneally on days 5 and 7. D, day; iv, intravenous; ip, intraperitoneal. (B) Tumor volume and body weight of the mice receiving treatment (n = 5 per group; *P < 0.05 and **P < 0.01). mAb, monoclonal antibody. (C) Tumor weight at end point of the study (n = 5 per group; *P < 0.05). (D) Concatenated dot plots showing percentages of CD4⁺ T cells and CD8⁺ T cells out of CD45⁺ leukocytes in the tumors. (E) Number of CD4⁺ FoxP3⁻ helper T cells and CD8⁺ T cells per gram of tumors (n = 5 per group; *P < 0.05 and **P < 0.01). (F) Ratio of CD8⁺ T cells and CD4⁺FoxP3⁺ regulatory T (T_reg) cells in tumors (n = 5 per group; *P < 0.05). (G) Quantification of the number of CD8⁺ T cells infiltrated to tumors by immunohistochemistry (n = 5 per group; *P < 0.05 and **P < 0.01).

To explore the immunological effect of the combinational ICB therapy, we examined the immunological hallmarks of T lymphocytes. The percentage of T lymphocytes within CD45⁺ leukocyte population within the tumors showed a 52.5, 85.7, and 105.1% increase in total CD8⁺ T cells for the vehicle LNPs + αPD-L1 αCTLA-4, T-RmRNA LNPs + isotype antibodies, and T-RmRNA LNPs + αPD-L1 αCTLA-4, respectively, compared with PBS control (Fig. 6D). The quantification of the total number of cells in tumors revealed a significant increase of CD4⁺FoxP3⁻ helper T cells (88.9% increase) and CD8⁺ cytotoxic T cells (262.6% increase) for the T-RmRNA LNPs + αPD-L1 αCTLA-4 group compared with the PBS group (Fig. 6E). We also observed that the T-RmRNA LNPs + isotype antibodies group showed a higher tumor infiltration of CD8⁺ T cells, which is consistent with our previous result for T-RmRNA LNPs (Fig. 5H). In addition, the ratio of CD8⁺ T cells to CD4⁺FoxP3⁺ regulatory T (T_reg) cells, representing the overall T cell activity, was improved when combined with PD-L1 and CTLA-4 antibodies (Fig. 6F). We also performed immunohistochemical staining and quantified CD8⁺ T cells within the tumors (fig. S9). Significantly more tumor-infiltrated CD8⁺ T cells were observed for the T-RmRNA LNPs + αPD-L1 αCTLA-4 group than those of the PBS and vehicle LNPs + αPD-L1 αCTLA-4 groups (Fig. 6G). These observations demonstrated that T-RmRNA LNPs + αPD-L1 αCTLA-4 treatment created a beneficial immunological environment with improved CD8⁺ T cell infiltration for T cell–mediated antitumor effect. Moreover, when ICB therapy was incorporated, CTLA-4 expression and CTLA-4⁺PD-1⁺ subsets of immunosuppressive T_reg cells were significantly down-regulated (fig. S10, A and B). This demonstrated that combining ICB alleviated the immunosuppressive effect of tumor-localized T_reg cells. Regarding the states of intratumor CD8⁺ T cells, the T-RmRNA LNP treatment groups (with or without αPD-L1 αCTLA-4) showed a trend toward an increase in CD62L⁻CD44⁺ effector T cells (fig. S10C). We also observed that PD-L1 and CTLA-4 antibody treatment up-regulated PD-1 expression in CD8⁺ T cells (fig. S10D). These findings collectively suggested that T-RmRNA LNPs + αPD-L1 αCTLA-4 treatment activated anticancer CD8⁺ T cells and triggered their functional exhaustion in the solid tumors (41).

We next examined the effector-memory phenotypes of CTLs in TdLNs following the combination treatment. Unlike CD8⁺ T cells in tumors, the cells in TdLNs were mainly composed of naïve (CD62L⁺CD44⁻) and central memory (CD62L⁺CD44⁺) T cell subsets (Fig. 7A), representing inactivated CD8⁺ T cells and stimulated CD8⁺ T cells circulated from tumors (42). We found that the T-RmRNA LNPs + αPD-L1 αCTLA-4 group significantly decreased naïve CD8⁺ T cells and increased central memory CD8⁺ T cells (Fig. 7B). This result demonstrates that the combination of LNPs and ICB treatment was able to establish antitumor memory immunity, which could contribute to an effective and long-term cancer immunotherapy.

Fig. 7. Central memory immunity of CTLs was up-regulated after combinational treatment of T-RmRNA LNPs and immune checkpoint inhibitors.

(A) Concatenated dot plots of CD62L and CD44 expression in CTLs (i.e., CD8⁺ T cells) of TdLNs with treatment of LNPs and anti–PD-L1 and anti–CTLA-4 or their isotype control antibodies. Gating was based on CD44 and CD62L expression of naïve cells in the spleen of a healthy mouse. Iso mAb, isotype monoclonal antibody. (B) Percentages of CD62L⁺CD44⁻ (naïve), CD62L⁺CD44⁺ (central memory), and CD62L⁻CD44⁺ (effector) subsets out of CD8⁺ T cells in TdLNs (n = 5 per group; *P < 0.05).

In addition, the toxicity and side effects of the LNPs, ICB antibodies, and combination treatments were assessed by hematoxylin and eosin (H&E) staining and histopathology examination on mouse organs with major distribution of LNPs, i.e., lung, liver, spleen, and kidney (figs. S11 and S12). Apart from extramedullary hemopoiesis and follicular lymphoid hyperplasia in spleens of all treatment groups, indicating increased systemic demand for granulocytes, erythrocytes, and lymphocytes during stimulation of immunity, no other significant morphological changes were observed in these organs of tumor-bearing mice following the treatment. This, and the minimal changes of mouse body weight that we see after treatment, reveals excellent tolerance and safety of our TAM-targeting LNPs.

DISCUSSION

TAMs are among the most abundant immune cells in TME, usually expressing immunosuppressive phenotypes that regulate the presence and function of antitumor immune cells, resulting in immune escape of tumor cells (43).

NPs modified with targeting moieties such as antibodies (44), peptides (45), and small molecules (46, 47) and carrying immunogenic payloads have been extensively explored to reeducate immunosuppressive TAMs for cancer therapy research. While most of these studies have reported substantially improved NP uptake by immunosuppressive TAMs, few studies have directly compared NP uptake between TAMs and other abundant cells, such as tumor cells, to demonstrate cell targeting. The lack of specificity and redundancy in cargo delivery not only compromises the efficiency of TAM modulation but may also induce undesired effects within tumors and increase toxicity.

In this study, we identified TREM2 as a TAM-specific target and demonstrated that TREM2 was exclusively expressed on immunosuppressive BMDMs as opposed to proinflammatory BMDMs and 4T1 tumor cells. TREM2 Ab-LNPs displayed altered in vitro cell uptake compared with unmodified LNPs, showing a TAM-to-tumor cell uptake ratio of 5.1. This is substantially higher than an earlier study reporting an NP modified with two peptides for targeting immunosuppressive macrophages and demonstrated 3.6-fold greater uptake by TAMs than B16F10 tumor cells at 24 hours post–intravenous injection (48). These findings highlight the potential of our Ab-LNPs to achieve more specific TAM targeting while minimizing off-target uptake by tumor cells for drug delivery.

Within the TME, chemokines and their receptors play an important role in mediating migration and recruitment of immune cells such as CD8⁺ T cells, reshaping immunotypes of tumors (e.g., from noninflamed to inflamed tumors) and affecting cancer progression (19). By accessing and analyzing TCGA data of patients with breast cancer, we identified that CXCL9 expression showed the most positive correlation with CD8 mRNA among all CXCL chemokines and the reported interleukin cytokines used for immunotherapy. This result is consistent with previous findings that CXCL9 acts as a chemoattractant of CD8⁺ T cells in tumors (49). To restore CXCL9 expression and investigate its antitumor effect, limited studies have reported findings using a transduced tumor model (50) and plasmid DNA-loaded NPs (51). Both approaches showed increased CXCL9 levels in tumors, resulting in enhanced CD8⁺ T cell infiltration. However, CXCL9 transduction is unlikely to be applicable in clinical settings. In the case of plasmid DNA-loaded NPs with no targeting moiety, the efficacy of CXCL9 expression was found to be highly tumor-type dependent. Notably, melanoma showed the highest expression level (51).

Here, we used mRNA-LNP technology for effective CXCL9 expression and incorporated TAM-targeting antibodies into LNPs. By reprogramming TAMs and delivering mRNA, we achieved highly efficient production of CXCL9 chemokine by TAMs in solid tumors. CXCL9⁺ macrophages have been reported to be positively correlated with tumor prognosis in clinical observations (52). Together, our approach focusing on CXCL9 mRNA and TAMs provides effective chemokine modulation within TME and holds potential for further investigation in clinical practice.

CTLs (i.e., CD8⁺ T cells) represent the most powerful antitumor effectors and form the foundation of cancer immunotherapy (4). However, the heterogeneous TME in solid tumors, particularly in breast cancer with noninflamed immunotypes and predominance of immunosuppressive cells, hinders tumor infiltration and antitumor activity of CD8⁺ T cells, thus limiting the response rate to clinically applicable immunotherapeutic strategies (53). To address these two obstacles, we developed TREM2 Ab-LNPs to codeliver the TLR7/8 agonist resiquimod and mRNA-CXCL9 to TAMs. Recent nanomedicine research has shown TLR7/8 agonists can act as a potent immunogenic adjuvant to NP delivery systems for macrophage reeducation, vaccination adjuvant, and combinational therapies (37, 54–56). However, systematic toxicities associated with TLR7/8 agonists remain a concern before clinical use; thus, it is essential to use lower doses or shorter courses to attain a beneficial immune response and avoid adverse effects (57). In our study, we show that low amounts of resiquimod-loaded LNPs not only can reprogram immunosuppressive TAMs but also can boost mRNA translation and CXCL9 production by the macrophages. Collectively, our LNP codelivery system addresses two critical steps of the immune cycle (T cell trafficking and immunosuppression of TME) (9) and creates a favorable immunological environment for CTLs.

As enhancing CTL activity can contribute to the response to clinically approved ICB treatment, we further investigated the combination of our LNPs with PD-L1 and CTLA-4 antibodies in mouse 4T1 tumors, a clinical strategy of combining immunotherapy. Our ICB combination therapy has demonstrated a further increase in the infiltration of tumor-localized helper T cells and CTLs, along with a decrease in T_reg cells. We found a significant increase in central memory CTLs in TdLNs, indicating the formation of antitumor memory immunity. However, the combinational treatment did not show any further enhancement of therapeutic efficacy. It is likely attributed to the lack of response to PD-L1 and CTLA-4 antibodies during our treatment course in 4T1 tumors. On the other hand, 4T1 tumors are known for a massive presence of myeloid-derived suppressor cells (58), which is yet to be addressed in our study. Further effort, thus, needs to be directed toward the optimization and combination with other therapeutics to further improve the treatment outcomes that we have seen.

One of the major concerns when using mRNA LNPs to express cytokines and chemokines is the potential toxicity and systemic inflammation caused to healthy organs. Therefore, the administration route must be carefully considered first. For instance, intratumoral injection has been used for LNPs carrying IL-12–encoding mRNA or circular RNA to activate natural killer and CD8⁺ T cells for anticancer therapies (59–61). Although effectiveness has been demonstrated, direct injection of LNPs into tumors limits their applicability across a wide range of tumor models and clinical settings. Furthermore, safety concerns are often dependent on the dosing amount of therapeutics and lipids, where effectiveness and toxicity need to be balanced. Our T-RmRNA LNPs consist of TAM-targeting antibodies and two immune modulators with an additive effect, directly improving the efficacy of CTL-mediated immunotherapy and, as a result, reducing the required dosing. We also demonstrated excellent tolerance of the LNPs and desired therapeutic and immunological effects via an intravenous administration route. Further reducing the dose of LNPs and ICB combinational therapy can still preserve the outcome. In addition, pathological examination of normal organs showed minimal side effects from T-RmRNA LNPs. While promising, a thorough investigation to evaluate systemic inflammation and long-term toxicity of the LNP treatment is necessary before proceeding to clinical trials.

Overall, our study offers a previously unreported Ab-LNP system that specifically delivers resiquimod and mRNA-CXCL9 to TAMs, relieving immunosuppression and expressing a CTL chemoattractant to overcome the immune barriers for CTL immunotherapy. The resultant LNPs achieved a moderate therapeutic efficacy while boosting tumor infiltration and activity of CTLs, further enhancing PD-L1 and CTLA-4 blockade therapy to create a CTL-favorable immunological TME and antitumor memory immunity. The efficacy and safety of the LNP system support further investigation and potential clinical translation for anticancer immunotherapy.

MATERIALS AND METHODS

TCGA analysis

Correlation between tumoral expression of CXCL chemokines; IL-12, IL-15, and CD8A mRNA levels; and survival probability of patients with breast cancer dependent on CXCL9 expression was analyzed using LinkedOmics platform (27). The platform uses and analyzes patient data from TCGA datasets. The results of gene correlations and prognosis of patients with breast cancer are in whole or part based on data generated by the TCGA Research Network: www.cancer.gov/tcga. Gene set enrichment analysis was also performed to identify the top 10 up-regulated and down-regulated pathways in coordination with CXCL9 expression (rank criteria, P value; minimal number of gene sets, 3; simulations, 1000). Correlation analysis between CXCL9 expression and the reported 68 stimulatory and inhibitory immune signatures (62) was performed, where genes with the top five correlation coefficients were highlighted, respectively.

Cell culture

4T1 cell line (RRID: CVCL_0125) was purchased from American Type Culture Collection. The 4T1 cells were grown in RPMI 1640 medium (Thermo Fisher Scientific, no. 11875093) supplemented with 10% fetal bovine serum (FBS; Thermo Fisher Scientific, no. A3160902) and 1% penicillin-streptomycin (Thermo Fisher Scientific, no. 15140122). Mouse bone marrow cells were harvested from healthy BALB/c mice (Laboratory Animal Services) and isolated according to a standard procedure published by Ying et al. (32). To generate BMDMs, the bone marrow cells were cultured in Dulbecco’s modified Eagle’s medium (Thermo Fisher Scientific, no. 11995065) supplemented with 10% FBS, 1% penicillin-streptomycin, and macrophage colony-stimulating factor (50 ng/ml; PeproTech, no. 315-02) for 7 days. BMDMs were further treated with mIL-4 (10 ng/ml; PeproTech, no. 214-14) to generate immunosuppressive BMDMs; LPS (100 ng/ml; Thermo Fisher Scientific, no. 00-4976-93) and mIFN-γ (50 ng/ml; PeproTech, no. 315-05) to generate proinflammatory BMDMs.

Animals

Female BALB/c mice (7 to 9 weeks old) were purchased from Laboratory Animal Services (South Australia, Australia) and Ozgene ARC (Perth, Australia) and maintained under pathogen-free conditions at the Laboratory Animal Services facility, University of Adelaide. Animal ethics approval was provided from the Faculty of Sciences, Engineering and Technology Animal Ethics Committee, University of Adelaide (animal ethics approval number S/2022/070). 4T1 cells were grown into a confluent state and tested to be free of mycoplasma contamination (Lonza) before in vivo use. Female BALB/c mouse (9 to 11 weeks old) was subcutaneously injected with 1 million 4T1 cells in 100 μl of PBS at the fourth mammary fat pad. Volume size of tumor was measured by a caliper (tumor volume = 0.5 × length × width × width). When tumor volume of each mouse was above 25 mm³, the mice were relocated into different groups in which approximate average volume of tumor was kept. Treatment schedule and experimental endpoint were set as per purposes and aims of the experiment.

mRNA production

A single batch of mRNA-encoding enhanced CXCL9 (mRNA-CXCL9) was manufactured by BASE mRNA facility, University of Queensland, Australia, according to their standard procedures of production. In brief, gBlock template was used to generate polymerase chain reaction (PCR) product, followed by in vitro transcription to produce mRNA. Cap1 analog (TriLink) and N1-methyl-pseudouridine modification were also incorporated. An mRNA purity of 97.30% was obtained by ion-pair PATfix high-performance liquid chromatography (BIA Separations), and the size of mRNA was confirmed by TapeStation (Agilent). In addition, endotoxin levels of the purified mRNA were measured with an eEndosafe endotoxin testing kit (Charles River Laboratory) and an endotoxin level of <0.331 EU/mg was recorded.

LNP preparation

Generally, all types of LNPs were prepared by mixing an ethanol phase containing lipids and aqueous phase containing mRNA and resiquimod (InvivoGen, no. tlrl-r848-1) with an N/P ratio of 8. The ethanol phase was prepared by dissolving heptadecan-9-yl 8-((2-hydroxyethyl)(6-oxO-6-(undecyloxy) hexyl) amino) octanoate (SM-102; SINOPEG), 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC; SINOPEG), cholesterol (Sigma-Aldrich), 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG2k; Cayman), and 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[maleimide(polyethylene glycol)-2000] (DSPE-PEG2k-maleimide; Biopharma PEG) at a molar ratio of 50, 10, 38.5, 1.2, and 0.3%, respectively. Total lipid concentration is 2.68 mM. The aqueous phase is composed of mRNA-eGFP, FLuc or CXCL9 (BASE facility, the University of Queensland), and/or resiquimod dissolved in 50 mM citrate buffer (pH 3.0). The aqueous phase was added to the ethanol phase and mixed by repeated pipetting at least 20 times to generate LNP solution. The LNP solution were thereafter dialyzed in 1× PBS (100 mM) using a dialysis membrane tubing with a molecular weight cutoff (MWCO) of 10,000 daltons (Thermo Fisher Scientific) for 24 hours at 4°C. Mannose LNPs were prepared by replacing DSPE-PEG2k-maleimide with 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-(polyethyleneglycol)-mannose (DSPE-PEG2k-mannose; Biopharma PEG). Vehicle LNPs were formulated by the same preparation method without adding mRNA or resiquimod in the aqueous phase.

TREM2 antibody coating on LNPs

TREM2 antibody coating was achieved on the basis of a reported protocol with some modifications. For small-scale experiment, 100 μg of TREM2 antibody (R&D Systems, rat IgG2B, clone no. 237920) or its isotype control antibody (R&D Systems, rat IgG2B, clone no. 141945) was mixed with tris(2-carboxyethyl) phosphine hydrochloride (TCEP; Sigma-Aldrich) in 100 μl of PBS (final concentration is 2 mM) and incubated at 37°C for 1 hour. TCEP and other small-molecule product were removed by passing the antibody solution through Zeba spin desalting columns (Thermo Fisher Scientific). For large-scale production of reduced antibodies, centrifugal filter with a MWCO of 3 kDa (Millipore) was used to purify the antibodies. The reduced antibody was quantified by NanoDrop (Implen) and freshly prepared to coat LNPs.

Antibody-coated LNPs (Ab-LNPs) were prepared by directly incubating the processed antibody and LNPs (DSPE-PEG2k-maleimide) in PBS at room temperature (RT) for 1 hour. Molar ratios of antibody and maleimide were kept at 0.25:1, 0.5:1, 1:1, 2:1, and 4:1. The Ab-LNPs were purified by centrifugal filter with a MWCO of 100 kDa (Sigma-Aldrich, no. UFC9100) and washed once by adding an equal volume of PBS for another filtration. The concentrated Ab-LNPs were stored at 4°C and used within 3 days for the experiment. The flow-through antibodies were collected and quantified by NanoDrop (Implen). Antibody-coating efficiency was calculated by (amount of antibody added − amount of antibody in flow-through solution)/antibody added × 100%.

Characterization of LNPs

To calculate loading efficiencies and weight percentages of resiquimod and mRNA in the LNPs, the LNPs were dissociated by 1% Triton X-100 at 4°C for 30 min. High-performance liquid chromatography (Shimazu UFLCxr) with ultraviolet detection at 260 nm was used to measure the loading efficiency of resiquimod. Phase A, Milli-Q water; phase B, acetonitrile solution with 0.1% v/v trifluoroacetic acid; running speed, 1 ml/min; and running time, 40 min. Loading efficiency of mRNA was measured by a standard RiboGreen assay (Thermo Fisher Scientific, no. R11490).

Hydrodynamic size, polydispersity index, and zeta potential of the LNPs were analyzed by DLS (ZetaSizer Ultra, Malvern). Morphology and size of the LNPs were characterized by a transmission electron microscope (FEI Tecnai G2 Spirit).

In vitro cell uptake assay and mRNA delivery

The mRNA-eGFP LNPs, TREM2 antibody-coated mRNA-eGFP LNPs (T-mRNA LNPs), and mannose mRNA-eGFP LNPs (M-mRNA LNPs) were stained by adding DiD (Thermo Fisher Scientific) at an amount of 0.5 mol % of SM-102 to the ethanol phase during LNP preparation. Mouse bone marrow cells (1.2 million) were seeded on a six-well plate with a coverslip covering the bottom. The cells were cultured in the culture medium for 7 days and then polarized with mIL-4 for 24 hours. To evaluate cell uptake, the DiD-stained LNPs were put into the wells at an mRNA concentration of 0.5 μg/ml in cell culture medium. At time points of 1, 4, and 24 hours, the cells were stained with DAPI (5 μg/ml; Thermo Fisher Scientific) and fixed with 5% paraformaldehyde (Sigma-Aldrich). A fluorescence microscope (Olympus BX51) was used to image the cells under DAPI and APC channels. Four regions of interest (ROIs) per sample condition were randomly selected and processed by ImageJ to obtain their integrated DiD fluorescence intensity and number of nuclei. DiD intensity per cell in ROIs was thereafter calculated.

For evaluation of mRNA delivery efficacy, the mRNA-eGFP LNPs (without DiD staining) were also added to 4T1 cells and mIL-4–treated BMDMs (when cells are around 80% confluent in a 12-well plate) at an mRNA concentration of 0.5 μg/ml. After incubation for 24 hours, cell suspensions were obtained by 0.5% trypsin-EDTA (Thermo Fisher Scientific) or a cell scraper and immediately analyzed by flow cytometry (BD FACSCanto II).

Confocal microscopy for LNPs

To study the effect of resiquimod on mRNA LNP uptake and lysosomal entrapment, mouse bone marrow cells were isolated, seeded on a 35-mm glass bottom dish (ibidi), and polarized by mIL-4. DiD-stained mRNA-eGFP LNPs (mRNA LNPs), DiD-stained resiquimod mRNA-eGFP LNPs (RmRNA LNPs), and a mixture of the mRNA LNPs and resiquimod (mRNA LNPs + R) were added to the cells, where the mRNA concentration was 0.5 μg/ml. The concentration of resiquimod was kept the same for RmRNA LNPs and mRNA LNPs + R. After incubation for 4 and 24 hours, cells were stained with DAPI (5 μg/ml) and 1 μM LysoTracker Red (Thermo Fisher Scientific) and then fixed with 5% paraformaldehyde. Zeiss LSM980 super-resolution confocal microscope with Airyscan 2 was used to acquire images under fluorescent channels of DAPI, AF568, and APC. Colocalization area and mean fluorescence intensity were obtained by ZEISS ZEN lite software. Experiment for the mRNA LNPs was also performed on proinflammatory BMDMs (i.e., treated with LPS and IFN-γ) as a comparison to the immunosuppressive BMDMs (i.e., treated with mIL-4) .

In vitro ELISA

Cell medium control, RmRNA-CXCL9 LNPs, T-RmRNA-CXCL9 LNPs, and M-RmRNA-CXCL9 LNPs (mRNA-CXCL9 concentration was 0.5 μg/ml) were incubated with mIL-4–treated BMDMs on a 24-well plate. After 24 hours, the control and LNP-containing medium were replaced by fresh cell culture medium and incubated for another 48 hours. The cell culture medium samples were collected, centrifuged, and diluted 10 times with PBS. CXCL9 concentration of the diluted samples was quantified by the CXCL9/MIG Quantikine ELISA Kit (R&D Systems) with an Infinite 200Pro microplate reader (Tecan) according to the manufacturer’s protocol.

RNA extraction, cDNA synthesis, and qPCR

After treatment with stimulants and/or mRNA-CXCL9 LNPs, total RNAs of BMDMs and 4T1 cells were isolated using the TRIzol reagent (Invitrogen) according to the manufacturer’s protocol. RNA concentration and quality were assessed by NanoDrop (Implen). For detection of TREM2, ARG1, MRC1, NOS2, and IL-12B mRNA, 500 ng of RNA was reverse transcribed using PrimeScript RT-PCR kit (Takara Bio, no. RR036A), followed by quantitative PCR (qPCR) using a TB Green Premix Ex Taq kit (Takara Bio, no. RR420A) with CFX96 real-time PCR detection system (Bio-Rad) according to the manufacturer’s protocol. The average relative expression was calculated using the 2^−ΔΔCt method, with glyceraldehyde-3-phosphate dehydrogenase expression used to normalize the data.

In vivo biodistribution study

The mRNA-FLuc LNPs, TREM2 antibody-coated mRNA-FLuc LNPs (T-mRNA LNPs), and mannose-modified mRNA-FLuc LNPs (M-mRNA LNPs) were labeled with DiD according to the protocol described above. PBS was used as the control. Female BALB/c mice (9 to 11 weeks old) were injected with 1 million 4T1 cells into their left fourth mammary fat pads. When tumor volumes of the mice exceeded 200 mm³, PBS and the LNPs were intravenously injected into the mice with a dose of mRNA-FLuc at 0.45 mg/kg body weight (n = 3 per group). After 12 hours, d-luciferin (Revvity) was injected intraperitoneally into the mice at 150 mg/kg. Luminescence and fluorescence imaging of the whole animal was performed by an IVIS Lumina X5 imaging system (Revvity), with a 620-nm excitation filter and 670-nm emission filter used for fluorescence. Tumor tissues and major organs (brain, lung, heart, liver, spleen, kidney, and muscle) were excised and imaged. All imaging data were analyzed and quantified by Living Image Software (Revvity). To evaluate macrophage and tumor cell uptake of the LNPs in TME, the excised tumors were further processed by Collagenase D (Sigma-Aldrich, no. 11088858001) digestion and stained with antibodies (table S2) for flow cytometry analysis (BD FACSCanto II). Data were further analyzed and replotted by FlowJo v10.

Immunofluorescence

Immunofluorescence imaging was performed on 5-μm sections of tumor tissues from in vivo biodistribution study to evaluate uptake of the LNPs by macrophages. The tumor tissue sections were fixed by 10% neutral buffered formalin for 10 min, washed twice, then blocked, and permeabilized with 5% bovine serum albumin (BSA) and 0.3% Triton X-100 (Sigma-Aldrich) for 20 min at RT. The sections were incubated with rat anti-mouse F4/80 (Abcam, ab6640, 2 μg/ml) or rat IgG2b isotype control antibody in 1% BSA for 2 hours at RT, then washed three times, and stained with anti-rat Alexa Fluor488 (Abcam, ab150165; 4 μg/ml) for 1 hour. Subsequently, the sections were stained with DAPI (0.5 μg/ml) at RT for 2 min. A slide scanner, Zeiss Axio Scan.Z1, was used to image the whole-tumor-tissue sections. Six 400 μm–by–400 μm high-power field regions in each section (per mouse) were randomly selected and used to quantify percentages of colocalization area of cells labeled with DiD-stained LNPs (APC channel) and F4/80⁺ macrophages (FITC channel). Average percentages of these six images per mouse were plotted.

Cytotoxicity assay

Raw 264.7 cells were seeded into 96-well plates at a density of 10,000 cells per well. T-mRNA LNPs and T-RmRNA LNPs were prepared and added to the cells at final mRNA concentrations of 7.8125, 15.625, 31.25, 62.5, and 125 ng/ml. Each group contains five replicates, and cells treated with fresh medium were set as control group. After 24 hours of incubation, the cell culture medium was replaced by fresh cell medium containing cell proliferation reagent WST-1 (Roche). After 2 hours, absorbance at 450 nm was measured using a microplate reader. Cell viability was calculated by normalization by the readings of control group.

Therapeutic study of LNPs

BALB/c mice (9 to 11 weeks old) were inoculated subcutaneously with 1 million 4T1 cells at their left fourth mammary fat pads on day 0. For LNP treatment, mRNA-CXCL9 LNPs (mRNA LNPs), resiquimod mRNA-CXCL9 LNPs (RmRNA LNPs), and TREM2 antibody–coated resiquimod mRNA-CXCL9 LNPs (T-RmRNA LNPs) were prepared according to the previously described. On days 7, 11, and 14 after tumor inoculation, PBS control or LNPs carrying 12 μg of mRNA were injected intravenously to tail vein of the mice (n = 5 per group). Tumor volume and body weight were monitored. Mice were humanely killed before ulceration takes place. At 6 hours after the last dose of the treatment, TdLNs (left inguinal lymph nodes), tumors, and major organs (lung, liver, spleen, and kidney) were excised from the mice.

To evaluate CXCL9 amount in tumor, ∼20 mg of tumor tissue was homogenized by TissueLyser II (QIAGEN) in 400 μl of PBS for 2 min, then centrifuged at 5000g for 10 min. The supernatant was diluted four times with PBS and quantified by the mouse CXCL9/MIG Quantikine ELISA Kit according to the manufacturer’s protocol.

For characterization of immune cells in tumor, a single-cell suspension of tumor tissues was obtained by mashing through a 70-μm cell strainer (Miltenyi Biotec). Cells were centrifuged at 500 rcf for 5 min at 4°C, washed with fluorescence-activated cell sorting buffer (PBS, 2% heat-inactivated FBS, and 2 mM EDTA) twice, and stained with antibodies listed in table S3. Precision counting beads (BioLegend) were added to the sample for quantification of the absolute number of immune cells. The mixture suspension was characterized and analyzed using a BD Fortessa X-20 and FlowJo v10.

Mouse ELISpot assay was performed to assess activity and response of cells from TdLNs. Multiscreen 96-well plates [0.45 μm, hydrophobic PVDF membrane (Millipore)] were coated with IFN-γ monoclonal antibody (5 μg/ml; eBioscience, no. AN-18) overnight at 4°C. Fresh cells isolated from TdLNs were added to wells at 2 × 10⁵ or 2.5 × 10⁵ cells in 100 μl of RPMI cell medium and then stimulated with 1.1 μg of 4T1 cell lysate and 0.25 μg of resiquimod. Cell medium and Ultra-LEAF purified anti-mouse CD3ε (BioLegend) were used as negative and positive controls. After 48 hours at 37°C, 5% CO₂, cells were removed, and secreted IFN-γ was detected by adding IFN-γ monoclonal antibody (2 μg/ml), Biotin (eBioscience, no. R4-6A2), followed by horseradish peroxidase–streptavidin (DAKO). The plate was developed by AEC chromogen concentrate and substrate Buffer (BioLegend). AID ELISPOT reader (AID GmbH) was used to scan the plate and analyze data.

Combinational therapy with immune checkpoint inhibitors

Balb/c mice (9 to 11 weeks old) were inoculated subcutaneously with 1 million 4T1 cells at their left fourth mammary fat pad on day 0. On days 4, 6, and 8, PBS, vehicle LNPs (i.e., LNPs without resiquimod and mRNA encapsulated), or T-RmRNA LNPs carrying 4 μg of mRNA-CXCL9 were injected intravenously to tail vein of the mice. On days 5 and 7 at 24 hours post–LNP treatment, 100 μg of anti–PD-L1 (Bio X Cell, no. BE0101) and 100 μg of anti–CTLA-4 (Bio X Cell, no. BE0131) or 100 μg of rat IgG2b (Bio X Cell, no. BE0090) and 100 μg of polyclonal Syrian hamster IgG isotype controls (Bio X Cell, no. BE0087) were injected intraperitoneally to the mice (n = 5 per group). Mice were humanely killed before ulceration occurred. Tumor volume and body weight were monitored. On day 12, TdLN, tumors, and major organs (lung, liver, spleen, and kidney) were excised from the mice in a double-blind manner.

For the immune profile of lymphocytes, single-cell suspensions of tumor tissue and TdLN for each mouse were obtained, processed with Foxp3/transcription factor staining buffer set (eBioscience, no. 00-5523-00), and stained with antibodies listed in table S4. Immune cells in tumor tissues were quantified with the addition of precision counting beads. BD Fortessa X-20 and FlowJo v10 were used for immune flow cytometry and data analysis, respectively.

Immunohistochemical staining of tumor sections

Tumors from combinational therapy were embedded in optimal cutting temperature compound and cut into 5-μm sections. The tumor sections were fixed with methanol for 10 min at RT, washed with PBS, and then incubated with 3% H₂O₂ for 30 min. Blocking was performed with 5% goat serum and 0.1% Triton X-100 in PBS for 45 min at RT. Rat anti-mouse CD8 alpha antibody (Bio-Rad, no. MCA2694) in 5% goat serum at 1:100 dilution was added to the sections and incubated in a humidity chamber overnight at 4°C. The sections were subsequently stained with goat anti-rat HRP antibody (Thermo Fisher Scientific, no. A10549) at 2 μg/ml, followed by visualization using a DAB substrate kit (Vector Laboratories, no. SK-4100) and counterstaining with hematoxylin. Zeiss Axio Scan.Z1 digital slide scanner was used to perform whole-tumor-section scans at 20× objective. Four 400 μm–by–400 μm images per mouse tumor were randomly selected for counting CD8⁺ cells located within tumor islets. The images with a number of CD8⁺ cells close to the average number of the group were picked as representative images.

H&E staining of mouse organs

To determine the impact of the treatment on LNP-distributed organs, mouse lung, liver, spleen, and kidney from both the therapeutic study and the combinational therapy study were sectioned and subjected to H&E staining. Histopathological examination was performed by a veterinary pathologist.

Statistical analysis and reproducibility

Data analyses in this study were performed using GraphPad Prism version 10 software. Data were presented as means ± SEM. Spearman’s correlation was used to assess the correlation of gene expression using the TCGA dataset. Statistical significance in mean values between two groups was determined by nonparametric Mann-Whitney tests. For comparison that involves more than two groups, one-way and two-way analysis of variance (ANOVA) were used. P < 0.05 indicated statistical significance. Double-blinded procedures were used for animal studies.

Supplementary Materials

This PDF file includes:

Figs. S1 to S16

Tables S1 to S4